Pre-salt discoveries are among the most important in the world over the last decade and ultra-deep water offshore exploration targets became common in oil industry. In these cases, the water layer typically compounds 30–40% of the whole geophysical region of interest. The most successful numerical schemes to model seismic waves in these regions are based on ordinary staggered-grid schemes (OSG). Nevertheless, these formulations have the drawback of requiring higher memory storage, since they need to save both velocity and stress components. Recently, we presented equivalent staggered-grid schemes (ESG) that allow memory storage reduction and/or higher computational efficiency depending on the specific case. Specially, the acoustic ESG has lower memory requirement and is the most efficient one. In order to improve the efficiency and reduce memory storage requirement in offshore regions, the present paper develops an explicit direct acoustic-elastic coupling to solve anisotropic wave equations. A domain decomposition is used, where the geophysical medium is divided in two different subdomains, one acoustic, related to water layer, and the other elastic, related to rock layers, considered as vertical-transverse isotropic (VTI) media. The best computational efficiency is achieved by choosing a flat-horizontal interface to couple the two subdomains. The former subdomain is modeled using the acoustic ESG and the second subdomain is modeled using the OSG for VTI media. Both schemes have the same stability properties and the resulting coupling algorithm has the same stability and accuracy as the original OSG applied to the whole model. The proposed method applies to 3D case up to orthorhombic case. To validate the method, it was applied to the Hess VTI model with water layer extended to compound 35% of the model. The reduction observed was near to 25% in memory and in CPU time and grows with percentage of water in the model.